Research Interests

Introduction

The main focus of my research is monitoring vegetation using Earth Observation methods. I have built a focus, in my own work and that of my research group, around developing and applying new methods in this area, with a key emphasis on moving the field from loose empirical correlations to making use of physically-based models. I have always sought to develop a strong, rigorous, capability in modelling within my own work and that of my group (research as well as training elements such as the MSc remote sensing). This has led to a wide range of fruitful collaborations over the years. The approach has provided practical solutions to operational monitoring (e.g. my work over the last 10+ years leading to the development of the NASA MODIS BRDF/albedo product with colleagues at BU¹, NASA, UoW, Swansea², and UCL) as well as providing for greater accuracy in spinoff applications such as model-based burn scar detection in Africa/change detection with colleagues at UMD³/NASA. Gaining a NERC EOSI Lectureship in the last few years (Dr Paul Saich, a microwave EO modeller) has allowed for rapid progress in the area of examining common links between information in the optical and microwave domains. Exploring and exploiting this synergy forms previous joint work for ESA and BNSC (in collaboration with BAE systems), as well as a focus for future developments in the NERC Centre of Excellence and other areas.

Developing fundamental understanding of radiation interactions

The basic approach of (robust) physically-based modelling developed has, and will continue to form a large component of my research focus. This has mainly been associated with developing models for 3D radiative transfer, of which our current model, drat, is an expression. This model has been applied to a wide range of canopy types, initially with the aim of developing a testbed simulation system, from which can be developed an understanding of the main factors affecting shortwave radiation scattering and absorption. We have participated with this model in model intercomparison exercises within the international framework of RAMI (Pinty et al., 2004).

The main directions of this work in the near future are (i) developing an understanding of the temporal dynamics of the radiation signal and synergy between optical and microwave scattered radiation through 3D dynamic modelling (see below), mainly through collaborations with INRA⁴ with ESA/BNSC/NERC funding; (ii) developing 'lumped parameter' models which encapsulate our understqanding of 3D scattering but which are able to accurately describe the radiation scattered with a small number of compound terms (see below); (iii) developing rapid methods for the inversion of physically-based models through sparse, compressed look-up tables (LUTs); (iv) developing system simulation tools to explore the possibilites of new sensor types and configurations.

Lumped parameter models

All of the streams of research in fundamental modelling are leading to the conclusion that whilst, in many ways, the remote sensing signal is a complex function of the details of canopy spatial arrangement and spectral properties, a decoupling of the structural and spectral aspects of scattering can provide significant insights into how a robust but simple model of radiation scattering and absorption can be constructed. Early evidence along these lines led to the development of so-called semi-empirical linear kernel-driven BRDF models which we helped developed and test and which form the basis of the NASA MODIS BRDF/albedo algorithm, a 16-day global 1 km product ( Schaaf et al., 2001). A spin-off application from these models has resulted in a set of 'signal tracking' models, one of which now forms the core of the NASA MODIS burned area detection algorithm. This product, which is soon to become operational, developed in collaboration with David Roy at UMD/NASA GSFC, allows the most accurate and robust mapping of burned areas so far developed (Roy et al., 2001, 2005). It has great potential for increasing our understanding of the impact of fire on Carbon budgets as well as in other areas (see fire below). Rebelo et al. (2005) show that semi-empirical signal tracking methods further developed from these models with an empirical temporal model allow for a more robust consideration of sudden changes in the reflectance signal, as well as more robust parameter mapping.

Whilst these semi-empirical models have allowed for a range of practical applications, the model parameters associated with them have no direct physical meaning. The 3D modelling work undertaken within my group (developing from Lewis & Disney, 1998 in Lewis et al., 2005; Disney et al., 2005) is pointing to the conclusion that if we separate the 'structural' aspects of scattering from the 'spectral' we can build much more meaningful 'lumped parameter' models, from which we can provide mappings of biophysical parameters, as well as more directly linking radiation scattering and absorption. This latter point is of particular importance if we are to use such models to drive vegetation/ecosystem growth models.

Monitoring and modelling photosynthesis from Space

A core stream of research within the CTCD is to develop improved modelling and monitoring of photosynthesis from space. We are doing this in several ways, using both simple models of Gross Primary Production (GPP) such as the Production Efficiency Approach (which more easily links to satellite observations) and more complex mechanistic models. This is currently developing in two main areas: (i) allowing for satellite observation-based estimates of GPP to be compared with mechanistic model predictions (Quaife et al., 2005); and (ii) allowing for the assimilation of satellite observations into mechanistic models to improve predictive capabilities and show model inadequacies. We are currently attempting this latter approach by assimilating satellite-derived products such as MODIS LAI, but we are also developing towards more directly linking the mechanistic model predictions to the observations through an 'EO operator', essentially a canopy scattering and absorption model driven by the mechanistic model state variables. This will use what we have learnt about simple parameterisatiuons of canopy scattering models and will allow for a clearer tracking of the uncertainties involved (thence uncertainties in Carbon flux prediction).

In addition, we are investigating new methods for measuring photosynthetic activity from space with methods ranging from simple band ratios (Photochemical Reflectance Index - PRI, which is linked to light-use efficiency through Xanthrophyll cycling), and more recently, passive detection of fluorescence.

Fire

Fire plays a major role in many ecosystem dynamics and is a large, dynamic and uncertain element of global Carbon bugets. One way in which Carbon release is estimated is by monitoring the area burned each year and multiplying this by factors associated with fuel load and fire intensity to arrive at estimates of Carbon released. We have therefore been working for several years to provide improved estimates of burned area from satellite data, building on our modelling expertise. As noted above, this has led to the development of the MODIS burned area product. An alternative and independent method for estimating Carbon release from biomass burning has been developed by Wooster (KCL). We have recently set up collaborative links with Wooster to combine these efforts with our improved burned area estimates and fire modelling activities within CTCD. We will be seeking NERC QUEST funding to follow this up.

Other

I have supervised around 20 MSc dissertations at UCL, and am currently responsible for four PhD students, three of whom are funded by NERC. I currently have three post-doctoral Research Fellows working for me on various NERC projects. I am involved in reviewing NERC and NASA grant applications, the former as a member of the NERC peer-review college, as well as reviewing papers for various journals including: RSE, RSR, IEEE GRS, IJRS, and New Phytologist.

1: BU = Boston University
2: UoW Swansea = University of Wales, Swansea
3: UMD = University of Marlyland
4: INRA = Institut Nationale de la Recherche Agronomique
5: CTCD = NERC Centre for Terrestrial Carbon Dynamics